Artificial Intelligence in Finance
Executive Summary
Key Takeaways
- ✓ Artificial intelligence in finance is not one technology but a family of distinct techniques, machine learning, natural language processing, and generative AI, each suited to different finance tasks and carrying different reliability characteristics.
- ✓ Machine learning is best suited to prediction and classification tasks over structured, historical data, forecasting, anomaly detection, credit scoring, where a large enough labelled dataset exists.
- ✓ Generative AI is best suited to language and reasoning tasks, drafting, summarisation, structuring unstructured commentary, rather than to producing verified numerical output on its own.
- ✓ The finance function has adopted AI unevenly by task — forecasting and anomaly detection have the longest track record of machine learning use; generative AI adoption in modelling and reporting is more recent and still maturing.
- ✓ Distinguishing which category of AI technique underlies a given tool or claim is a prerequisite for judging that tool's reliability for a specific finance task, addressed throughout this domain.
Objective¶
This guide maps how artificial intelligence is applied across the finance function today, as the entry point to AI Financial Modelling & Artificial Intelligence in Finance.
Three Categories of AI Technique¶
Machine learning learns statistical patterns from historical, structured data to predict or classify a future or unseen value, a revenue forecast, a credit default probability, an anomalous transaction. It requires a sufficiently large, representative, labelled dataset and produces output whose reliability can be measured empirically against held-out data. See Machine Learning.
Natural language processing (NLP) extracts structure and meaning from unstructured text, contract terms, earnings call transcripts, footnote disclosures, converting language into data a downstream process can use. See Natural Language Processing.
Generative AI produces new language or content in response to a prompt, drafting commentary, summarising a document, or restructuring input, using a large language model's learned patterns rather than a fixed rule set. See Large Language Model and Generative AI.
These are not competing technologies; a single finance workflow often uses more than one in sequence, NLP to extract data from a contract, machine learning to score a resulting risk, and generative AI to draft the resulting memo.
Where Each Technique Fits in the Finance Function¶
Financial modelling and FP&A. Machine learning supports forecasting and driver identification; generative AI supports scenario drafting and narrative commentary. See AI in Financial Modelling.
Risk and audit. Machine learning supports anomaly detection and credit scoring; NLP supports contract and disclosure review. See Financial Model Auditing for how deterministic, rule-based methodology differs from either.
Treasury and reporting. Generative AI supports drafting board and investor commentary from structured outputs, with human review remaining the control that verifies the underlying numbers.
Adoption Maturity by Function¶
Forecasting and anomaly detection have the longest track record of machine learning use in finance, often a decade or more of production deployment in larger institutions. Generative AI adoption in modelling support, drafting, and reporting is considerably more recent, and adoption maturity varies significantly by institution and function, addressed in full in AI Adoption Framework.
Common Construction Pitfalls¶
Treating "AI" as a single undifferentiated technology. A tool described only as "AI-powered" gives no information about whether it is applying machine learning, NLP, or generative AI, each of which carries different reliability characteristics for a given task.
Applying generative AI where machine learning is the better fit. Using a language model to attempt numerical prediction, rather than a structured forecasting method, forgoes the empirical accuracy measurement machine learning affords.
Assuming uniform adoption maturity across finance functions. Treating generative AI's recency in modelling and reporting as equivalent to machine learning's decade-plus track record in forecasting overstates how proven a given application actually is.
Recommended Practices¶
- Identify explicitly which category of AI technique, machine learning, NLP, or generative AI, underlies a given tool or claim before assessing its suitability for a task.
- Match the technique to the task: prediction and classification to machine learning, language and drafting to generative AI, unstructured text extraction to NLP.
- Treat adoption maturity as a function-specific and technique-specific question, not a single finance-wide answer.
Continue Reading¶
Related Pillars¶
Related Technical Guides¶
Related Glossary¶
How OXXON tests thisRun a free structural check with FMAE
Frequently Asked Questions
What is artificial intelligence in finance?
The application of AI techniques, principally machine learning, natural language processing, and generative AI, across finance functions including financial modelling, FP&A, risk management, treasury, and audit, to support prediction, classification, drafting, and analysis tasks.
What are the main categories of AI technique used in finance?
Machine learning (prediction and classification over structured data), natural language processing (extracting structure and meaning from text), and generative AI (producing new language or content from a prompt), each with distinct strengths and reliability characteristics.
Which finance tasks is machine learning best suited to?
Prediction and classification tasks over structured, historical data where a sufficiently large labelled dataset exists, forecasting, anomaly detection, and credit scoring, being the most established examples.
Which finance tasks is generative AI best suited to?
Language and reasoning tasks, drafting commentary, summarising documents, and structuring unstructured input, rather than producing verified numerical output on its own, which remains the domain of a structured financial model.
Has AI adoption been even across finance functions?
No. Forecasting and anomaly detection have the longest track record of machine learning use. Generative AI adoption in modelling support and reporting is more recent and still maturing, with adoption maturity addressed in the AI Adoption Framework guide.
Why does it matter which category of AI underlies a specific finance tool?
Because machine learning, NLP, and generative AI carry materially different reliability properties, a tool's suitability for a given task depends on which category of technique it actually uses, not on how advanced the tool is described as being.
References
Related Articles
AI Financial Modelling & Artificial Intelligence in Finance
AI financial modelling is the application of machine learning and generative AI techniques within the financial modelling process itself, driver identification, construction assistance, scenario generation, and narrative drafting, while artificial intelligence in finance is the broader application of those same technique categories across the finance function generally. This page is the hub for the Knowledge Centre's AI financial modelling content: the foundational distinction between machine learning, natural language processing, and generative AI; how AI accelerates modelling construction without replacing the auditable calculation layer beneath it; a staged framework for adopting AI reliably; enterprise applications across FP&A, forecasting, valuation, and investment analysis; governance and risk practice; and the institutional best practice synthesis this domain builds toward.
AI in Financial Modelling
AI in financial modelling refers to the application of machine learning and generative AI techniques within the modelling process itself, rather than across the finance function broadly: identifying candidate drivers from historical data, assisting with formula and structure construction, generating scenario variations, and drafting narrative commentary around a model's output. This guide sets out where these applications add genuine value and, just as importantly, where the calculated number itself must remain the output of a structured, auditable model rather than of the AI directly.
Machine Learning
Machine learning is a category of artificial intelligence technique that learns statistical patterns from historical, structured data in order to predict or classify a future or unseen value. In finance, it underlies forecasting, anomaly detection, and credit scoring applications, and its reliability is established empirically, by measuring predictive accuracy against held-out historical data, rather than by auditing a fixed rule set.
Generative AI
Generative AI is a category of artificial intelligence technique, most commonly a large language model, that produces new language or content, text, summaries, drafted formulas, in response to a prompt. In finance, it is well suited to drafting, summarisation, and narrative tasks, and is distinct from machine learning, which predicts or classifies from structured historical data rather than generating new content.
Natural Language Processing
Natural language processing, or NLP, is the category of artificial intelligence technique that extracts structure and meaning from unstructured text, contract terms, earnings call transcripts, footnote disclosures, converting language into data a downstream process can use. In finance, NLP typically feeds structured data into machine learning or a financial model, functioning as an input stage rather than a decision-making or generative stage on its own.
AI Adoption Framework
AI adoption in a finance function is most reliable when treated as a staged progression rather than an immediate wholesale rollout: exploratory pilots on low-stakes tasks, supervised production use on defined tasks with human checkpoints, and a fully governed operating model with defined ownership and controls. This guide sets out each stage, the specific conditions an organisation should meet before advancing, and why skipping stages tends to produce ungoverned, inconsistent adoption rather than faster value capture.